Communication Prediction via Neural Networks of the Residual Hydrogen Peroxide used in Photo- Fenton Processes for Effluent Treatment This communication proposes the use of neural networks in the prediction of resi- dual concentrations of hydrogen peroxide from the treatment of effluents through Advanced Oxidative Processes (AOP’s), in particular, the photo-Fenton process. To verify the efficiency of the oxidative process, the Chemical Oxygen Demand (COD) parameter, the values of which may be modified by the presence of oxidiz- ing agents such as residual hydrogen peroxide, is frequently taken in account. The analysis of the H 2 O 2 interference was performed by spectrophotometry at 450 nm wavelength, via the monitoring of the reaction of ammonia with metavanadate. The results of the hydrogen peroxide residual concentration were modeled via a feedforward neural network, with the correlation coefficients between actual and predicted values above 0.96, indicating good prediction capacity. Keywords: Hydrogen peroxide, Neural networks, Photo-Fenton Received: March 20, 2007; revised: May 6, 2007; accepted: May 7, 2007 DOI: 10.1002/ceat.200700113 1 Introduction Several effluents are not degraded by biological or physical- chemical treatments. Thus, it is necessary to employ other tech- niques, such as Advanced Oxidative Processes (AOP’s), which may lead to the degradation of a wide range of organic com- pounds, with the formation of CO 2 ,H 2 O and minerals. Ozone and/or hydrogen peroxide may be used in the AOP’s with or without ultraviolet radiation. Iron salts can also be utilized with hydrogen peroxide (Fenton reagent) or combined with hydro- gen peroxide and UV light, i.e., the photo-Fenton process [1]. Such processes are classified as homogeneous. AOP’s can also be classified as heterogeneous processes, with the use, for instance, of titanium dioxide together with solar or ultraviolet radiation. The oxidative process is characterized as a nonlinear process, where interactions occur among the various input variables. These features result in the problem having a remarkable mathe- matical complexity. In relation to the number of input and out- put variables, the process modeling is characterized by multi- dimensional modeling. In this context, neural networks are presented as a technique capable of modeling the AOP’s, since they represent nonlinear phenomena without the development of the complex mathematics involved in such processes. In recent years, neural networks have been applied to several different chemical engineering fields, e.g., with respect to AOP’s, Pareek et al. [2] used a feedforward type neural network to study the photo-degradation of spent Bayer liquor. Pearson correla- tion coefficients above 0.99 were obtained. Slokar et al. [3] uti- lized Konen type neural networks to model the Reactive Red 120 dye decoloration due to the use of H 2 O 2 /UV. Salari et al. [4] also applied the neural model technique in treating waters contami- nated with methyl tert-butyl ether (MTBE) through the use of hydrogen peroxide and ultraviolet radiation. In this work, the authors utilized a feedforward network to predict the MTBE concentration after photo-oxidizing treatment, obtaining Pear- son correlation coefficients equal to 0.998. In particular, with regard to the photo-Fenton process, the work of Durán et al. [5], which involved the neural simulation of the Reactive Blue 4 dye degradation process, can be men- tioned. In this work, the constants of kinetics of decoloration and mineralization were modeled as output variables with mean errors below 18 % and 14 %, respectively. 2 Photo-Fenton Process The Fenton reagent is a mixture of hydrogen peroxide and iron salts that produces the hydroxyl radical, which is a strong oxi- dant, through the reaction given by Eq. (1) 1) : © 2007 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim http://www.cet-journal.com Oswaldo L. C. Guimarães 1 Henrique Otávio Queiroz de Aquino 1 Ivy S. Oliveira 1 Darcy Nunes Villela Filho 1 Helcio José Izario Filho 1 Adriano Francisco Siqueira 1 Messias Borges Silva 1,2 1 Escola de Engenharia de Lorena – Universidade de São Paulo – USP, Sao Paulo, Brasil. 2 School of Engineering of Guaratinguetá – Sao Paulo State University – UNESP, Sao Paulo, Brasil. – Correspondence: O. L. C. Guimarães (oswaldocobra@debas.eel.usp.br), Escola de Engenharia de Lorena – Universidade de São Paulo – USP, Lorena, São Paulo, Rodovia Itajubá-Lorena Km 74.5, CEP 12.602-810, Brasil. – 1) List of symbols at the end of the paper. 1134 Chem. Eng. Technol. 2007, 30, No. 8, 1134–1139